This document presents a neural network and fuzzy logic based protection scheme for series compensated double circuit transmission lines. It aims to overcome challenges from series compensation, mutual coupling between lines, and fault conditions like resistance. The proposed scheme uses a neural network to estimate the power system condition based on local measurements. Fuzzy logic is then used to classify fault types and determine appropriate tripping decisions. Simulation results show the scheme can accurately detect and classify different fault types under varying system conditions like compensation level and fault location. The adaptive nature of this neural network and fuzzy logic approach provides robust protection for series compensated double circuit transmission lines.
IRJET- Design of Four Stack Millimeter Antenna for Medical Applications
If3115451551
1. Bhawna Nidhi, Ramesh Kumar, Amrita sinha / International Journal of Engineering Research
and Applications (IJERA) ISSN: 2248-9622 www.ijera.com
Vol. 3, Issue 1, January -February 2013, pp.1545-1551
Neural Network and Fuzzy Logic Based Protection of Series
Compensated Double Circuit Transmission line
Bhawna Nidhi, Ramesh Kumar, Amrita sinha
(Electrical Engineering Department, M.Tech. student, NIT Patna, India)
(Electrical Engineering Department, Associate Professor, NIT Patna, India)
(Electrical Engineering Department, Professor, NSIT Patna, India)
ABSTRACT
Main purpose of every relay is to detect mutual coupling impedance between double circuit
and classify the fault and isolate the faulty part lines and also by the values of the fault resistance
as soon as possible. Protective distance relays, [4]. Series compensation along with its protective
which make use of impedance measurements in devices can undermine the effectiveness of many of
order to determine the presence and the protection schemes used for long distance
identification of faults, are not accurate to that transmission lines [8-9]. In order to overcome the
extent after installed series capacitance on the problems related to series compensation as well as
line. In this proposed scheme, Neural network parallel lines we need an adaptive and robust
and Fuzzy logic are proposed for protection of system.
series compensated double circuit transmission Successful application of Adaptive Neural
line. A Neural Network is used to estimate the and fuzzy systems in electrical power system has
actual power system condition that improves the demonstrated that this powerful tool can be
protection system selectivity. Fuzzy logic is used employed as an alternative method for solving
for decision making that offers an appropriate problems accurately and efficiently. The features of
tripping decision. The effect of series ANN such as ability to learn, generalize and
compensation, mutual zero-sequence coupling, robustness of fuzzy system are combined to make
remote infeed and fault resistance have been more efficient system. Several researches carried
considered. out for fault analysis of single circuit transmission
line using ANN [5-7]. Srivani has given adaptive
Keywords - mutual coupling, series protection method for series compensated single
compensation, adaptive protection, artificial neural line with double end fed [8]. However, in this paper
network, fuzzy logic effects of mutual coupling are missing. Pradhan has
proposed a positive sequence directional relaying
I. INTRODUCTION scheme for single line series compensated circuit
Now-a-days demand of reliable and but has not considered mutual coupling [9]. Thoke
smooth power supply is going to increase due to proposed protection method using ANN for double
use of electronic devices in industries and in home circuit transmission line but series compensation
appliances. To meet this high demand, power has not considered in his paper [10]. Kale and
system has become more complicated so the Gracia has also given the methods for double line
protection has become challenging task for the protection without series compensation [7].
engineers [1]. Fixed series compensation and Makwana proposed method for protection of
parallel lines are the ways commonly used for double circuit transmission line with series
better utilization of transmission system [2-3]. Such compensation by using symmetrical component
type of transmission system requires more advance theory [11].Parikh in his work used SVM for fault
security control to deliver smooth power supply. classification in series compensated line.[12]
Performance of the conventional digital distance In this paper an adaptive scheme using
relay is affected by the presence of zero sequence ANN and Fuzzy logic is proposed for protection of
transmission line. Different faults are introduced in
transmission line then pre-fault and post-fault data
1545 | P a g e
2. Bhawna Nidhi, Ramesh Kumar, Amrita sinha / International Journal of Engineering Research
and Applications (IJERA) ISSN: 2248-9622 www.ijera.com
Vol. 3, Issue 1, January -February 2013, pp.1545-1551
Fig-1 Simulated circuit of double circuit series compensated network
are collected for proposed method. The influence of Table- 1.The power system network simulation has
the power system condition is summarized as given been performed with MATLAB-SIMULINK.
in the Table-2. Finally, the results of the application
of concepts are presented in this paper. Table1-Parameters of Parallel Transmission Line
The aim of this paper is in two fold. Firstly, it Positive sequence resistance R1, 0.01809
is shown that ANN compensates the actual power ῼ/KM
system condition as compensation level, fault Zero sequence resistance R0, ῼ/KM 0.2188
resistances, inception angle, fault location, which Zero sequence mutual resistance , 0.20052
influences the protection system. ANN is also R0m ῼ/KM
capable of estimating the general power system Positive sequence inductance L1, 0.00092974
condition by using local measurement. Secondly, H/KM
fuzzy logic expresses complex system linguistically Zero sequence inductance L0, H/KM 0.0032829
by using IF, THEN condition. Fuzzy logic system Zero sequence mutual inductance, 0.0020802
used to discriminate different types of faults and L0m, H/KM
phase involved, so that make trip decision accurate. Positive sequence capacitance C1, 1.2571e-008
This method is robust and requires very less time in F/KM
detection and classification of faults without using Zero sequence capacitance C0, F/KM 7.8555e-009
remote end data. Zero sequence mutual capacitance -2.0444e-009
C0m, F/KM
II. POWER SYSTEM NETWORK Length 300
SIMULATION
A three-phase, 50 Hz, 735 kV power
system transmitting power from a six 350 MVA
generators to an equivalent network through a 300
km transmission line has been considered as shown
in Fig-1. Both lines are series compensated indicate
in block diagram. For each line, each phase of the
series compensation module contains the series
capacitor, a metal oxide varistor (MOV), protecting
the capacitor, and a parallel gap protecting the
MOV. When the energy dissipated in the MOV
exceeds a threshold level of 30 MJ, the gap
simulated by a circuit breaker is fired. Parameters Fig 2 . Proposed protection scheme
of double circuit Transmission line are shown in
1546 | P a g e
3. Bhawna Nidhi, Ramesh Kumar, Amrita sinha / International Journal of Engineering Research
and Applications (IJERA) ISSN: 2248-9622 www.ijera.com
Vol. 3, Issue 1, January -February 2013, pp.1545-1551
Simulation Result
Typical examples of different types of simulated faults are shown in figure 3
Figure3- Current Waveform Of Different Types Of Fault
1547 | P a g e
4. Bhawna Nidhi, Ramesh Kumar, Amrita sinha / International Journal of Engineering Research
and Applications (IJERA) ISSN: 2248-9622 www.ijera.com
Vol. 3, Issue 1, January -February 2013, pp.1545-1551
III. RELAY MODELING BASED ON NEURAL Table 2 Pattern Generation
NETWORK AND FUZZY LOGIC
Major functional blocks of proposed relay are SL PARAMETER SET VALUE
shown in figure-4 NO
1 % of compensation 30,40,50,60,70
2 Fault Location(KM) 20,40,80,120,200,24
0,280
3 Fault Inception angle 0 & 90 deg
4 Fault Resistance 0,50 and 100 ohm
5 Fault type a1g,a2g,b1g,b2g,c1g,c2g
,a1b1,a2b2,a1c1,a2c2,b1
c1,b2c2,a1b1g,a2b2g,a1
c1g,a2c2g,b2c2g,b1c1g,
Fig-4.Process for Generating Input Patterns to the a1b1c1,a2b2c2
Relay
Total 3520 samples including no fault data has
been used for training for ANN.10 % of data is
3.1 Neural Network Modelling used for testing which chosen arbitrary and
The Neural network model has designed to be remaining 90% used for training
used in distributed and learning based environment. The fully connected three-layer feed-
The architecture provides learning from data and forward neural network (FFNN) was used to
approximate reasoning. classify faulty/non-faulty data sets and the error
back-propagation algorithm was used for training.
3.1.1 Input Layer One hidden layer was found to be adequate for the
The neurons of input layer represent the fault classification application. Hyperbolic tangent
input variable. Here currents of six phases, the function was used as the activation function of the
phase voltages and zero and negative sequence hidden layer neurons. Pure linear function was used
currents are given as input to the neural network. for the output layer. Training is done with Back
The current (I) and voltage (v) signals are Propagation algorithm and Levenberg-Marquardt
processed so as to simulate a 4 kHz sampling (LM) algorithm.
process (80 samples per 50Hz cycle) .This The input layer simply transfers the input
sampling rate is compatible with sampling rates vector x, to the hidden neurons. The outputs uj of
presently used in digital relays for protection the hidden layer with the sigmoid activation
process. function and transferred to the output layer, which
DFT is used for calculating the is composed of seven neurons. The output value of
fundamental components of currents and voltages the neuron in the output layer with the linear
of different phases. Negative and zero sequence are activation function gives the phase of the
calculated from sequence analyzer block of transmission line in which fault occur1 (the
MATLAB toolbox. presence of a fault) or 0 (the non-faulty situation).
Input X= After training, the neural fault detector was tested
[Ia1,Ib1,Ic1,Ia2,Ib2,Ic2,Ineg1,Izero1,Ineg2,Izero2, with 90 new fault conditions and different power
Va, Vb, Vc] system data for each type of fault are given in
Table 3-4.
3.1.2 Output Layers
Output layer of ANN consist 7 neurons. Each
output gives values corresponding to six phase of
double line and ground.
Output Y= [A1, B1, C1, A2, B2, C2, G]
3.1.3 Training set
Training set of neural network has to
accurately represent the problem. For Preparing
the training set that represents cases the ANN needs
to learn pre fault and post fault data are processed.
These data are generated by varying different
parameter of power system. Parameters that are
vary during data collection are shown in table-2 Figure-4 Fuzzy System
1548 | P a g e
5. Bhawna Nidhi, Ramesh Kumar, Amrita sinha / International Journal of Engineering Research
and Applications (IJERA) ISSN: 2248-9622 www.ijera.com
Vol. 3, Issue 1, January -February 2013, pp.1545-1551
been represented as linguistic variable low and
high. The output of fuzzy logic module has been
Table- 3:.Change In Fault Location (Km) defined by linguistic variable as normal and types
Inception Angle 0, Compensation 60%, Fault of faults in series compensated double line as
Resistance=0 shown in table- 5.The trapezoidal membership
function is used for input and triangular for output.
outp Fault=a1g fault=ab1
ut D=20 D=20 D=28 D=20 D=2 D=280 While the output data from Fuzzy unit are:
0 0 00 1 ≤ output <0.5 for no fault (No Trip)
-
a1 0.999 0.987 0.92 0.999 0.996 0.9961 0.5≤ output <1.5 for a1g fault ( Trip line one)
2 1 output < 2.5 for b1g fault (Trip line one)
.5≤
b1 2.0e- 0.009 0.031 0.999 0.999 0.99 2.5≤ output <3.5 for c1g fault (Trip line one)
04 3 9 3.5≤ output < 4.5 for ab1 fault (Trip line one)
c1 0.003 3.4e- 0.00 0.026 6.4e- 3.2e-05. 4.5≤ output <5.5 for ac1 fault (Trip line one)
4 04 05 5.5≤ output < 6.5 for bc1 fault (Trip line one)
a2 0.026 0.005 0.016 0.000 1.1e- 0.0086 6 output <7.5 for ab1g fault (Trip line one)
.5≤
1 2 05 7. ≤ output < 8.5 for ac1g fault (Trip line one)
5
b2 0.016 0.002 0.007 0.000 0.008 0.027 8 output <9.5 for bc1g fault (Trip line one)
.5≤
7 7 9.5 output<10.5 for abc1 fault (Trip line one)
≤
c2 0.042 0.018 0.00 3.7e- 0.032 2.1e-04 10.5≤output 11.5for a2g fault (Trip line Two)
<
3 6 06 6 11.5≤output<12.5 for b2g fault (Trip line Two)
g 0.989 0.999 0.934 1.0e- 0.003 .0034 12.5≤ output <13.5 for c2g fault (Trip line Two)
05 2 13.5≤ output < 14.5 for ab2 fault (Trip line
Two)
14.5≤ output <15.5 for ac2 fault (Trip line Two)
Table-4: Change In Level Of Compensation
15.5≤ output <16.5 for bc2 fault (Trip line Two)
Inception angle=0, Distance=20Km, fault
16.5≤ output<17.5 for ab2g fault (Trip line
resistance=0
Two)
Outpu Fault=a2g Fault=ac2g
17.5≤ output<18.5 for ac2g fault (Trip line
t 30% 50% 70 30% 50% 70%
Two)
%
18.5≤ output<19.5 for bc2g fault (Trip line
a1 0.02 0.11 0.20 1.10e- 0.000 0. Two)
4 7 0 05 098
19.5≤ output<20.5forabc2 fault (Trip line Two)
b1 0.14 0.00 0.02 0.005 5.59e- 0.026 IV. ANALYSIS OF RESULT
4 7 0 06
c1 0.07 0.00 0.16 6.48e- 0.000 0.413 4.1 Speed of proposed scheme
7 6 7 05 This system provides very high speed in detection
a2 0.97 0.95 0.74 0.996 0.997 0.716 and classification of different types of fault. It takes
2 0 2 only 6-9 sample of fault to detect its types.
b2 0.00 0.42 0.08 0.008 0.019 0.134
0 7 1
c2 0.00 0.07 0.01 0.832 0.981 0.80
4 0 7
G 0.99 0.79 0.95 0.999 0.794 0.977
7 1 7
3.2 Fuzzy logic based Decision making Module
Fuzzy logic Module has been designed to
translate the linguistic variable into symbolic terms.
The behavior of such systems is described through
a set of fuzzy rules, like:
IF <premise> THEN <consequent>
That uses linguistics variables with symbolic terms.
Each term represents a fuzzy set. The terms of the
input space (typically 5-7 for each linguistic
variable) compose the fuzzy partition.
The output of ANN has been used as input
variable to the Fuzzy system. The seven output
nodes of ANN module are converted into fuzzy
variables. The numerical output data of ANN has
1549 | P a g e
6. Bhawna Nidhi, Ramesh Kumar, Amrita sinha / International Journal of Engineering Research
and Applications (IJERA) ISSN: 2248-9622 www.ijera.com
Vol. 3, Issue 1, January -February 2013, pp.1545-1551
The adaptive protection scheme is tested under a
defined fault type, but for different compensation,
fault locations, fault resistances and fault inception
angles. All the test results clearly show that the
proposed adaptive protection technique is well
suited for series compensated double-circuit lines
with remote end infeed.
REFERENCES
1. P.M.Anderson, Power system protection.
New York: IEEE Press, 1999.
2. A.G.Phadke and Stanley H.horowitz,
Fig-5 Performance Of System With Sample
“power system relaying”
Number
3. P.M.Anderson, “Series compensation of
power system’. PBLSH Inc.Encinitas CA
4.2 Accurate Tripping
4. Y.Hu, D.Novosel, M.M.Saha and
Fuzzy logic module issue trip signal only
V.Leitloff, An adaptive scheme for
when it confirms that fault is of permanent type. In
parallel line distance protection , IEEE
the case of transient fault or no fault condition no
Trans. Power Del. Vol.17 no1,pp.105-
trip signal is issued. This module is used for
110,jan 2002
decision making and to prevent mal-operation of
5. Bouthiba ,Tahar(2004) ,Fault location in
the relay. Response of fuzzy system for different
EHV transmission line using artificial
condition shown in table-5
neural
networks,Int.J.Appl.Math.Compt.Sci.Vol-
4.3 Decisions of fault type
14 ,no.1,pp.69-78
The proposed system can also distinguish
6. J. Gracia and A. J. Mazón and I. Zamora,
between transient fault, permanent fault and no
(2005) ,Best ANN Structures for Fault
fault condition. For differentiate between fault and
Location in Single and Double-Circuit
transient condition we compare some sample of
Transmission Lines, IEEE transactions on
output, here 5-10th sample of output used for this.
power delivery, vol. 20, NO. 4,
Z=Yn-Yn-1 =0 No Fault
7. Kale ,V.S and Bhide,S.R and Bedekar,
Z=Yn-Yn-1 <0 Transient Fault
P.P. and G.V.K.Mohan (2008), Detection
Z=Yn-Yn-1 >0 Permanent Fault
and Classification of Faults on Parallel
Transmission Lines using Wavelet
From above analysis it is clear that the
Transform and Neural Network ,
trained Neural network with Fuzzy logic is robust
International Journal of Electrical and
to different parameter variations. Due to the fact
Computer Engineering , vol-3,pp-15
that this system is a nonlinear compensator which
8. Srivani, S.G.and Vittal, K.P.(2010),
adapts to the variation between faulted voltages,
Adaptive distance relaying scheme in
currents and fault positions under different fault
series compensated transmission lines,
resistance conditions.
IEEE Conference Publications Power
Electronics, Drives and Energy Systems
V. CONCLUSION: (PEDES) , , Page(s): 1 - 7
The influence of the mutual coupling of 9. Pradhan, A and Jena, P (2011), A positive
parallel circuit and impedance of series capacitor sequence directional relaying algorithm
on the accuracy of relays depends on actual power for series compensated line, IEEE Trans.
system condition. Power and Energy Society General
The proposed adaptive protection offers an Meeting, , Page(s): 1 – 1
approach to cope with the influence caused by 10. Jain, Anamika and Thoke A.S. (2009),
variable power system conditions .The selectivity Classification of single line to ground fault
of the protection system is increased, and so is the on double circuit transmission line using
power system reliability. It is shown that ANN can ANN, Int .journal of computer and
very well be used to construct the relation between electrical engineering, vol-1, no-2
local measurement and actual power system 11. Vijay H.Makwana and Bhavesh R.Bhalja
condition. Most importantly, sensitivity of the relay ,A New Digital Distance Relaying Scheme
to power system condition change is reduced to for Series-Compensated Double-Circuit
almost zero with Fuzzy logic decision making Line During Open conductor and Ground
system.
1550 | P a g e
7. Bhawna Nidhi, Ramesh Kumar, Amrita sinha / International Journal of Engineering Research
and Applications (IJERA) ISSN: 2248-9622 www.ijera.com
Vol. 3, Issue 1, January -February 2013, pp.1545-1551
Fault , IEEE Trans On power support vector machine. International
delivery,Vol-27,No-2,April-2012 jaournal of Electrical power and energy
12. Urmil B.Parikh ,Biswarup Das and system. Vol-32, page 629-636
Rudraprakash Maheshwari (2010) , Fault
classification technique for series
compensated transmission line using
Table5 Fuzzy logic based control unit output
a1 b1 c1 a2 b2 c2 g Fuzzy Trip
output
1.0000 0.1225 0.9965 0.0024 2.1e-05 0.0482 2.4e-04 0.93 1
1.0000 1.0000 0.4105 0.0036 0.0047 1.4e-08 6.8e-04 1.2 1
1.0000 0.0943 0.0200 1.6e-05 1.07e-04 5.7e-06 1.0000 0.83 1
0.0103 2.4e-06 9.7e-07 1.0000 0.8243 0.0671 0.0904 1.87 2
4.1e-06 2.1e-04 2.9e-07 1.0000 0.0045 0.0031 0.9165 1.63 2
9.8e-06 6.0e-06 2.1e-07 1.0000 0.0045 0.9382 0.9147 1.97 2
0.0019 0.0854 0.0072 2.4e-06 1.0000 1.0000 0.2494 2.3 2
4.7e-04 0.0038 0.9999 0.1342 1.7e-07 1.6e-07 1.0000 0.69 1
0.0341 0.0023 0.0089 0.0263 0.0752 0.0863 0.0097 -0.32 0
0.0045 0.2452 0.3187 0.0031 0.4231 0.3473 0.0213 0.27 0
1.0000 0.9999 0.9999 4.7e-06 3.3e-06 0.0100 0.0011 1.45 1
1551 | P a g e